Chemisist over on bitcointalk has added an optimization to to latest mining code.
I am getting about a 30% boost from it on my core 2 machines and from what people have been saying it's about the same on core i's. Amd FX processors seem to mine slower however, my FX 8120 definitely did.
I would be careful about building his source, he like to make frequent commits without testing. Version bf0ff8fa5593896d800661d288dcfdc7c8a96747 works well.
Another small optimization I have notices people using is to mine with more threads than processor cores. 1.5-2x threads/cores seems to work well this can be done with "setgenerate true 6" (for 6 threads) in the debugger or "rpcthreads=6" in the config file.